System Modeling and Application in Gearbox Fault Diagnosis Based on EMD and ARX Model

The EMD method as a means of signal pre-processing is used to decompose the vibration signal of gearbox into a number of IMF components, using the largest energy component as an effective component of the output response signal and the input shaft rotate speed or torque signal as the system input signal, then the time-series ARX model is established, in this paper. Extracted model parameters are as a feature vector of recognized working state of gearbox, and as input parameters of RBF neural network to achieve automatic recognition of faults in the gearbox. The results show that: ARX model can enhance the anti-jamming capability of model, and have great significance for improving the accuracy of fault diagnosis by calculating the fluctuations of input signal into the model.